Exploring attributes

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(readr)
library(data.table)

## Warning: package 'data.table' was built under R version 3.2.5

## 
## Attaching package: 'data.table'

## The following objects are masked from 'package:dplyr':
## 
##     between, last

library(tidyr)
library(knitr)
library(ggplot2)
library(ggrepel)
library(broom)
library(reshape2)

## 
## Attaching package: 'reshape2'

## The following objects are masked from 'package:data.table':
## 
##     dcast, melt

combinewithoutuser<-read_csv("combinewithoutuser_complete.csv")

business<-readRDS("yelp_academic_dataset_business.rds")

business_simple<-business%>%filter(grepl("Restautrant|Food|Breakfast|Bars|food|Burgers|Brunch|Sandwiches|Pubs|Chinese|Italian|American|Pizza|Coffee|Tea|Fast Food|Asian|Fusion|Lounges|Cafes|Irish|Gluten|Salad|Diners|Seafood|Bakeries|Desserts|Japanese|Ice Cream & Frozen Yogurt|Tapas/Small Plates|Mediterranean|Wine Bars|Vegetarian|Portuguese|German|Delis|Chicken Wings|Hot Dogs|Polish|Greek|Sushi Bars|Indian|Mexican|Bagels|Donuts|Tapas Bars|Cocktail Bars|Ethnic Food|Middle Eastern|Steakhouses|Cafeteria|Candy Stores|Korean|Chocolatiers & Shops|Cheese Shops|Vietnamese|Thai|Tea Rooms|Latin American|Creperies|French|Taiwanese|Buffets|Cajun/Creole|Soul Food|Juice Bars & Smoothies|Fondue|Ethiopian|Persian/Iranian|Popcorn Shops|Spanish|Cheesesteaks|Fish & Chips|British|Kosher|Armenian|Cupcakes|Vegan|Hawaiian|Cuban|Gastropubs|Russian|Pretzels|Fruits & Veggies|Gelato|Halal|Dim Sum|Filipino|Pasta Shops|Mongolian|Colombian|Cantonese|Street Vendors|Belgian|Cambodian|Hungarian|Szechuan|Bubble Tea|Laotian|African|Beer Bar|Himalayan/Nepalese|Moroccan|Falafel|Indonesian|Turkish|Afghan|Food Stands|Modern European|Irish Pub|Brazilian|Food Court|Malaysian|Coffeeshops|Hot Pot|Burmese|Macarons|Ramen|Empanadas|Bistros|Teppanyaki|Brasseries|Singaporean|Champagne Bars|Scandinavian|Canadian|Poutineries|Haitian|Arabian|Austrian|Czech|Slovakian|Bangladeshi|Egyptian|Dominican|Scottish|Patisserie/Cake Shop|Pub Food|Puerto Rican|Australian|Ukrainian|Sri Lankan|Beer Garden|International|Beer Gardens|Serbo Croatian|Kebab|Alsatian|Oriental|Shanghainese|Venezuelan|Bavarian|Iberian|Curry Sausage|Rhinelandian|Beer Hall|Eastern European|Wok|Trinidadian|Swiss Food|Pita",categories))

business_simple<-business_simple%>%
  filter(city%in%c("Pittsburgh","Charlotte","Urbana","Champaign","Phoenix", "Scottsdale","Tempe","Mesa", "Chandler","Gilbert","Glendale", "Las Vegas","Henderson", "Madison"))
business_simple$state[business_simple$business_id=="g49oTp73Pk_WpOfQVtmcew"] <- "NV"

## Plotting the distribution of review_counts per stars
combinewithoutuser %>% ggplot(aes(x = stars)) + geom_bar()

combinewithoutuser %>% summarize(mean = mean(stars), var = var(stars)) ## not Poisson

## Source: local data frame [1 x 2]
## 
##      mean       var
##     (dbl)     (dbl)
## 1 3.54474 0.4090287

#library(gridExtra)
#grid.arrange(map_nv1,map_nv2, nrow=1)

#all_states <- map_data("state")
#ggplot() + geom_polygon( data=all_states, aes(x=long, y=lat, group = group),colour="white", fill="grey10" ) + geom_point(data=restaurant_loc, aes(longitude, latitude), col="red", cex=3) + geom_point(data=station_loc, aes(lon, lat), col="white", cex=2) + facet_wrap(~state)

Restaurant Locations

The longitudes and latitudes are used to map all restaurant locations in our dataset with R GIS (Geographical Information System). The size of the dots reflects the number of restaurants in a specific area. The bigger the dot, the larger the number of restaurants in the area.

## Distribution of all restaurants in our dataset
business_simple%>%group_by(city)%>%ggplot( aes(city) ) + geom_bar( stat="count", width = 0.8, fill="darkolivegreen3")

business_simple%>%group_by(state)%>%ggplot( aes(state) ) + geom_bar( stat="count", width = 0.8, fill="darkolivegreen3")

restaurant_loc<-read_csv("restaurant_loc.csv")
station_loc<-read_csv("station_loc.csv")
restaurant_loc<-combinewithoutuser%>%left_join(restaurant_loc)

## Joining by: "business_id"

restaurant_loc2<-restaurant_loc%>%filter(review_count > 100)

library(RgoogleMaps)

## Warning: package 'RgoogleMaps' was built under R version 3.2.5

PlotOnStaticMap(lat = restaurant_loc$latitude, lon = restaurant_loc$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=1.4, pch=19, col="red3", FUN = points, add = F)

## NV
state<-restaurant_loc%>%filter(state=="NV")
map_nv1<-PlotOnStaticMap(lat = state$latitude, lon = state$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

state2<-restaurant_loc2%>%filter(state=="NV")
map_nv1<-PlotOnStaticMap(lat = state2$latitude, lon = state2$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

## PA
state<-restaurant_loc%>%filter(state=="PA")
map_nv1<-PlotOnStaticMap(lat = state$latitude, lon = state$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

state2<-restaurant_loc2%>%filter(state=="PA")
map_nv1<-PlotOnStaticMap(lat = state2$latitude, lon = state2$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

## NC
state<-restaurant_loc%>%filter(state=="NC")
map_nv1<-PlotOnStaticMap(lat = state$latitude, lon = state$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

state2<-restaurant_loc2%>%filter(state=="NC")
map_nv1<-PlotOnStaticMap(lat = state2$latitude, lon = state2$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

## IL
state<-restaurant_loc%>%filter(state=="IL")
map_nv1<-PlotOnStaticMap(lat = state$latitude, lon = state$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

state2<-restaurant_loc2%>%filter(state=="IL")
map_nv1<-PlotOnStaticMap(lat = state2$latitude, lon = state2$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

## AZ
state<-restaurant_loc%>%filter(state=="AZ")
map_nv1<-PlotOnStaticMap(lat = state$latitude, lon = state$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

state2<-restaurant_loc2%>%filter(state=="AZ")
map_nv1<-PlotOnStaticMap(lat = state2$latitude, lon = state2$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

## WI
state<-restaurant_loc%>%filter(state=="WI")
map_nv1<-PlotOnStaticMap(lat = state$latitude, lon = state$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

state2<-restaurant_loc2%>%filter(state=="WI")
map_nv1<-PlotOnStaticMap(lat = state2$latitude, lon = state2$longitude,  zoom =5, size = c(640,640), TrueProj=TRUE, cex=0.7, pch=19, col="red3", FUN = points, add = F)

Distribution of variables

## Type of restaurant
combinewithoutuser %>% ggplot(aes(type)) + geom_bar( stat="count", width = 0.8, fill="darkolivegreen3")

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(type)) + geom_bar( stat="count", width = 0.8, fill="darkolivegreen3") +
  scale_y_log10() +
  theme(axis.text.x = element_text(angle = 40, size = 10, hjust = 0.9, vjust = 0.9)) +
  facet_wrap(~state)

## City
combinewithoutuser %>% ggplot(aes(city)) + geom_bar( stat="count", width = 0.8, fill="darkolivegreen3")

## State
combinewithoutuser %>% ggplot(aes(state)) + geom_bar( stat="count", width = 0.8, fill="darkolivegreen3")

## Review_count
combinewithoutuser %>% group_by(city) %>% 
  ggplot(aes(city, review_count, col=city)) + geom_boxplot() +
  ggtitle("Review_count, stratified by city")

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, review_count, col=state)) + geom_boxplot() +
  ggtitle("Review_count, stratified by state")

## Review_count < 1000
combinewithoutuser %>% filter(review_count < 1000) %>% group_by(city) %>% 
  ggplot(aes(city, review_count, col=city)) + geom_boxplot() +
  ggtitle("Review_count, stratified by city")

combinewithoutuser %>% filter(review_count < 1000) %>% group_by(state) %>% 
  ggplot(aes(state, review_count, col=state)) + geom_boxplot() +
  ggtitle("Review_count, stratified by state")

## Stars
combinewithoutuser %>% group_by(city) %>% 
  ggplot(aes(city, stars, col=city)) + geom_boxplot() +
  ggtitle("Review_count, stratified by city")

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, stars, col=state)) + geom_boxplot() +
  ggtitle("Review_count, stratified by state")

## Stars vs review_count
combinewithoutuser %>% group_by(stars) %>% 
  ggplot(aes(stars, review_count, col=stars, group=stars)) + geom_boxplot() +
  scale_y_log10() +
  ggtitle("Review_count, stratified by stars") + ylab("Review_count (log10 scale)")

Distribution of binary attributes

test <- combinewithoutuser %>% select(business_id,type,city,state,review_count,stars,takeout:goodforbreakfast,alcohol,price:noise,wifi)

## Overall averages of attributes
r_sum <- summary(test)
r_sum_df <- as.data.frame(r_sum[4,])
colnames(r_sum_df) <- "Average"
r_s <- as.data.frame(cbind (rownames(r_sum_df), substr(r_sum_df$Average, 9,16)))
names(r_s) <- c("Attribute", "Average")
r_s[-1:-4,] %>% kable

	Attribute	Average
5	review_count	113.3
6	stars	3.545
7	takeout	0.9421
8	reservation	0.3368
9	outdoorseating	0.4512
10	waiterservice	0.6105
11	creditcards	0.9818
12	goodforkids	0.8323
13	goodforgroups	0.9197
14	goodfordessert	0.03297
15	goodforlatenight	0.06838
16	goodforlunch	0.4446
17	goodfordinner	0.3071
18	goodforbrunch	0.09984
19	goodforbreakfast	0.1117
20	alcohol	0.5349
21	price	1.594
22	parking	0.9392
23	noise	0.9197
24	wifi	0.4032

## plotting of binary attributes
r_s_binary<-r_s[-1:-6,]
r_s_binary<-r_s_binary[-15:-18,]
r_s_binary %>% ggplot( aes(Attribute, Average) ) + geom_bar( stat="identity", position = "stack", width=0.9, fill="darkseagreen") + coord_flip(xlim=c(0,15), ylim=c(0,20)) + geom_text(aes(label=Average), hjust=1, color="white", size=3.5) + theme_minimal() + geom_hline(aes(yintercept=14), color="red")

## plotting of categorical attributes
r_s_cat<-r_s[21:24,]
r_s_cat$hline <- c(4,2,3,2)
r_s_cat %>% ggplot(aes(Attribute, Average)) + geom_bar( stat="identity", position = "stack", width=0.9, fill="darkseagreen") + coord_flip(xlim=c(0,5), ylim=c(0,10)) + geom_errorbar(aes(y=hline, ymax=hline, ymin=hline), colour="#AA0000") + geom_text(aes(label=Average), hjust=1, color="white", size=3.5) + theme_minimal()

## Averages of attributes by state
test_pa <- test %>% filter(state=="PA")
r_sum <- summary(test_pa)
r_sum_df <- as.data.frame(r_sum[4,])
colnames(r_sum_df) <- "Average"
r_s <- as.data.frame(cbind (rownames(r_sum_df), substr(r_sum_df$Average, 9,16)))
names(r_s) <- c("Attribute", "Average")
r_s[-1:-3,] %>% kable

	Attribute	Average
4	state	NA
5	review_count	71.90
6	stars	3.63
7	takeout	0.9282
8	reservation	0.3942
9	outdoorseating	0.3625
10	waiterservice	0.7032
11	creditcards	0.9161
12	goodforkids	0.7153
13	goodforgroups	0.8504
14	goodfordessert	0.0146
15	goodforlatenight	0.07178
16	goodforlunch	0.4173
17	goodfordinner	0.4282
18	goodforbrunch	0.07664
19	goodforbreakfast	0.07543
20	alcohol	0.4939
21	price	1.687
22	parking	1.464
23	noise	0.9294
24	wifi	0.3771

## plotting of binary attributes
r_s_binary<-r_s[-1:-6,]
r_s_binary<-r_s_binary[-15:-18,]
r_s_binary %>% ggplot( aes(Attribute, Average) ) + geom_bar( stat="identity", position = "stack", width=0.9, fill="darkseagreen") + coord_flip(xlim=c(0,15), ylim=c(0,20)) + geom_text(aes(label=Average), hjust=1, color="white", size=3.5) + theme_minimal() + geom_hline(aes(yintercept=14), color="red")

## plotting of categorical attributes
r_s_cat<-r_s[21:24,]
r_s_cat$hline <- c(4,2,3,2)
r_s_cat %>% ggplot(aes(Attribute, Average)) + geom_bar( stat="identity", position = "stack", width=0.9, fill="darkseagreen") + coord_flip(xlim=c(0,5), ylim=c(0,10)) + geom_errorbar(aes(y=hline, ymax=hline, ymin=hline), colour="#AA0000") + geom_text(aes(label=Average), hjust=1, color="white", size=3.5) + theme_minimal()

Distribution of binary attributes by state

test <- combinewithoutuser %>% select(business_id,type,city,state,review_count,stars,takeout:goodforbreakfast,alcohol,price:noise,wifi)
r_sum_bin <- test%>%group_by(state)%>%select(state,takeout:alcohol)%>%summarise_each(funs(mean))

mdat <- melt(r_sum_bin, id.vars="state")
head(mdat)

##   state variable     value
## 1    AZ  takeout 0.9674077
## 2    IL  takeout 0.9636364
## 3    NC  takeout 0.9597378
## 4    NV  takeout 0.9095637
## 5    PA  takeout 0.9282238
## 6    WI  takeout 0.9081197

ggplot(mdat, aes(variable, value, fill=state)) + 
  geom_bar(stat="identity", position="dodge") + 
  geom_hline(yintercept=1, col="red", cex=0.5) +
  theme(axis.text.x = element_text(angle = 40, size = 10, hjust = 0.9, vjust = 0.9)) +
  ggtitle("Average attributes by state") + scale_fill_brewer() + 
  theme(panel.background = element_rect(fill = "black")) +
  theme(plot.background = element_blank() 
        ,panel.grid.major = element_blank() 
        ,panel.border = element_blank())

Distribution of categorical variables

r_sum_cat <- test%>%group_by(state)%>%select(state,stars,price:wifi)%>%summarise_each(funs(mean))

mdat <- melt(r_sum_cat, id.vars="state")
head(mdat)

##   state variable    value
## 1    AZ    stars 3.551263
## 2    IL    stars 3.454545
## 3    NC    stars 3.541199
## 4    NV    stars 3.516325
## 5    PA    stars 3.630170
## 6    WI    stars 3.580128

ggplot(mdat, aes(variable, value, fill=state)) + 
  geom_bar(stat="identity", position="dodge") + 
  theme(axis.text.x = element_text(angle = 40, size = 10, hjust = 0.9, vjust = 0.9)) +
  ggtitle("Average attributes by state") + scale_fill_brewer() + 
  theme(panel.background = element_rect(fill = "black")) +
  theme(plot.background = element_blank() 
        ,panel.grid.major = element_blank() 
        ,panel.border = element_blank())

test %>% ggplot(aes(stars)) + geom_histogram( stat="bin", bins=17, fill="lightskyblue")

test %>% ggplot(aes(price)) + geom_histogram( stat="bin", bins=10, fill="lightskyblue")

test %>% ggplot(aes(parking)) + geom_histogram( stat="bin", bins=10, fill="lightskyblue")

test %>% ggplot(aes(noise)) + geom_histogram( stat="bin", bins=10, fill="lightskyblue")

test %>% ggplot(aes(wifi)) + geom_histogram( stat="bin", bins=10, fill="lightskyblue")

Distribution of demographics

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, population_zip, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, income_zip, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, age_zip, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, white, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, Black.or.African.American, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, American.Indian.and.Alaska.Native, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, Asian, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, Native.Hawaiian.and.Other.Pacific.Islander, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, other, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, LANDSQMI, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, popdensity_zip, col=state)) + geom_boxplot()

combinewithoutuser %>% group_by(state) %>% 
  ggplot(aes(state, education_zip, col=state)) + geom_boxplot()